If Writing is Thinking…

The page argues that deep reading is rare in business settings, even for long memos, PRDs, budgets, proposals, and deal docs. It says people usually skim for context or confirmatory details, and that org memos are one of the few formats that get read at all, often only when they include visuals like an org chart.

It then asks what changes when AI does the writing and the author does not fully understand the result. It extends the concern to summarized text that may lose information or invent data, and connects the same reading problem to science and Wall Street, where reviewers, analysts, and readers often do not read full documents carefully.

Reading notes

#

• In business environments, almost no one fully reads what is written, even when the material is long and important.
• The author’s own experience with long memos suggested that only a few people would read them end-to-end.
• Slides were easier to use in meetings, but the memo still had to be revisited afterward because people had not absorbed it.
• Organization-wide memos were the most reliably read documents, especially when they included an org chart image.
• The page treats this as a problem for large companies as well as for AI, since AI can write text that no one, including the author, has read carefully.
• AI summaries can be lossy or invent data, which raises another layer of risk.
• The same reading behavior appears in science, where reviewers often scan papers for specific issues instead of reading them fully.
• The page links the reproducibility crisis partly to that lack of careful reading.
• It also points to Wall Street, where long analytical write-ups are consumed more for narrative confirmation than for full verification.